steps: - launch mail (imap, will fetch mail once) - put mac to sleep - wake up - click "get mail" spinning watch, mac goes into la la land at: (CurStackBase does not seem to apply...dumping 4K.) Calling chain using A6/R1 links Back chain ISA Caller 0D229B40 PPC 0EFB1F78 PRP_TryLock+00994 0D229AD0 PPC 0EEC3990 nsThread::Main(void*)+00028 0D229A80 PPC 0CEDD838 nsSocketTransportService::Run()+0019C 0D229A10 PPC 0CEDD038 nsSocketTransportService::ProcessWorkQ()+00074 0D2299C0 PPC 0CED8850 nsSocketTransport::Process(short)+0058C 0D229950 PPC 0CED9620 nsSocketTransport::doWrite(short)+00030 0D229900 PPC 0CED97F8 nsSocketTransport::doWriteFromBuffer(unsigned int*)+00068 0D2298B0 PPC 0EEDB3C8 nsPipe::nsPipeInputStream::ReadSegments(unsigned int (*)(nsIInpu tStream*, void*, const char*, unsigned int, unsigned int, unsigned int*), void*, unsigned int , unsigned int*)+0013C 0D229820 PPC 0CED93C8 nsWriteToSocket(nsIInputStream*, void*, const char*, unsigned in t, unsigned int, unsigned int*)+00034 0D2297E0 PPC 0EFAEC98 PR_Write+00030 0D2297A0 PPC 0EFAC5B0 PR_SetThreadDumpProc+00184 0D229760 PPC 0EFA87EC PR_GetConnectStatus+00660 Closing log Recall that every mac ships with the default of going to sleep after 30 minutes. Every mac owner who uses mail will encounter this bug. It is not fixed by Gordon's fixes to the DNS service for browser.
this is a dup of a bug gordon is working on already where nspr loses all it's connections and it hangs when we try to right into a closed connection. Gordon, can you mark this a dup of thatproblem you are working on?
well, this is different, since navigator now works with his patch, and we probably won't be able to fix this problem by RTM (i hope we can get it to stop crashing, but i doubt a real fix will be allowed in). i'm inclined to keep these two bugs separate, though gordon should own them both, i guess. also cc'ing gagan, so he is up on what's going on.
Assignee: mscott → gordon
Yes, let's keep these bugs separate.
Ok, Mike and I have a potential fix by commenting out two lines in macsockotpt.c. I need to do a bit more investigation to understand why I put the two lines of code there in the first place (a year and a half ago), and have Netscape Steve Dagley and sfraser review the change. The result is the app doesn't lock up any more, and mail works again after closing and reopening the window (no need to quit and relaunch the app).
The two lines we are commenting out are this case in SendReceiveStream(): // case kOTOutStateErr: // it has been closed // return 0; I would have provided a diff, but I have other changes waiting for rtm++ in this file. See bug 49990.
putting on the right radar.
Whiteboard: [rtm need info]
Okay, I've read through the Open Transport documentation again, and talked with Wan-Teh about the semantics of the SendReceiveStream() function. I am convinced this is the right fix now. I'll get a super-review and check it into the trunk, according to Gagan's request.
Status: NEW → ASSIGNED
Whiteboard: [rtm need info] → [rtm+]
sr=sfraser. I suggest moving the case kOTOutStateErr: to down above the default: case with fallthrough; the logic will be the same.
yeah, that sounds better than commenting out the lines. good idea.
Whiteboard: [rtm+] → [rtm++]
The fix has been landed on the Netscape_20000922_BRANCH and NSPRPUB_CLIENT_BRANCH. I will land the fix on the nspr tip and NSPRPUB_RELEASE_4_0_BRANCH in the near future.
The future is now. Fix checked into nspr tip and NSPRPUB_RELEASE_4_0_BRANCH.
Status: ASSIGNED → RESOLVED
Closed: 19 years ago
Resolution: --- → FIXED
Verfied OK using 2000-10-18-08 mn6 branch commercial build, mac OS 9.0. Note to trunk verifier: Before fix, my machine hit this lockup every time I tried, whereas a coworker's machine didn't. You may want to verify on a couple machines...
verified with 110606 mozilla trunk build on OS9. I went back to a 10/2/00 build and sure enough it locked me up. I tried with today's trunk build and it didn't lock me up. Setting bug status to Verfied and removign vtrunk keyword.
Status: RESOLVED → VERIFIED
You need to log in before you can comment on or make changes to this bug.